btrfs: fix RAID direct I/O reads with alternate csums
authorOmar Sandoval <osandov@fb.com>
Mon, 2 Mar 2020 22:02:49 +0000 (14:02 -0800)
committerDavid Sterba <dsterba@suse.com>
Tue, 3 Mar 2020 14:26:08 +0000 (15:26 +0100)
commite7a04894c766daa4248cb736efee93550f2d5872
tree500ed1c5e3a81c03d1c70c7c3f3e44e476a154e5
parenta5ae50dea9111db63d30d700766dd5509602f7ad
btrfs: fix RAID direct I/O reads with alternate csums

btrfs_lookup_and_bind_dio_csum() does pointer arithmetic which assumes
32-bit checksums. If using a larger checksum, this leads to spurious
failures when a direct I/O read crosses a stripe. This is easy
to reproduce:

  # mkfs.btrfs -f --checksum blake2 -d raid0 /dev/vdc /dev/vdd
  ...
  # mount /dev/vdc /mnt
  # cd /mnt
  # dd if=/dev/urandom of=foo bs=1M count=1 status=none
  # dd if=foo of=/dev/null bs=1M iflag=direct status=none
  dd: error reading 'foo': Input/output error
  # dmesg | tail -1
  [  135.821568] BTRFS warning (device vdc): csum failed root 5 ino 257 off 421888 ...

Fix it by using the actual checksum size.

Fixes: 1e25a2e3ca0d ("btrfs: don't assume ordered sums to be 4 bytes")
CC: stable@vger.kernel.org # 5.4+
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/inode.c